Read Aloud the Text Content

This audio was created by Woord's Text to Speech service by content creators from all around the world.

Text Content or SSML code:

Social mining is generally understood as representing, analysing and extracting enforceable trends and patterns from raw data in social media. The current research aimed at combining both visual and textual characteristics to identify bullying posts in social media. This paper has introduced a novel framework to identify Cyberbullying instances with the new integrated representation of image and text. This important contribution provides an analytical background that opens the way to combine different forms of data to be trained in a single system instead of parallel systems where different systems are used for different types of data. Our proposed system can correctly identify 74% of the cases of bullying class. Overall, our system got 68% weighted average F1-score of both (bullying and non-bullying) classes. We found that a single layer of convolution with a larger filter size is better than multiple layers of convolution with a lesser number of filters. We have only considered the image and text for Cyberbullying detection task, but audio, video and URLs of the post can also be useful information that may be considered for identifying bullying scenarios. Finally, despite introducing a unified representation of different modalities, future research should aim to determine the proper weight of text and image into a Cyberbullying identification task.